01
OpenClaw plugin
Observes supported tool attempts and completions through the configured OpenClaw path.
Governance for action-taking AI agents
Know what your AI agents did, what policy says, and what still needs review.
McPherson Governance evaluates supported OpenClaw activity against explicit policy and produces a readable record of what was attempted, what completed, what policy would have decided, and what remains unknown.
The current release operates in shadow mode. It has no authority to block, approve, deny, or rewrite your agents’ actions.
The accountability problem
An activity log is not the same as governance evidence.
A tool event can be useful without answering the questions an operator, reviewer, or incident owner actually needs resolved.
- A log may not proveThe correct action occurred.
- A log may not provePolicy was followed.
- A log may not proveApproval was required or obtained.
- A log may not proveThe external action completed.
- A log may not proveThe evidence is sufficient.
- A log may not proveThe issue is resolved.
The product in three parts
Observation, policy, and evidence that stay honest about their limits.
02
Policy service
Evaluates mapped activity against explicit governance rules and records the decision the policy would have produced.
03
Observa
Turns governance evidence into an operator-readable record of what happened, what needs review, and what should not be assumed.
How the governed path works
A readable path from proposal to review.
Each stage has a distinct job. The current public release observes and records; it does not give remote policy authority over execution.
- Agent proposesA tool action enters the supported OpenClaw path.
- Plugin observesThe configured attempt is described with minimized metadata.
- Policy evaluatesMapped activity is checked against explicit rules.
- Evidence is boundAttempt, decision, and observed completion are correlated.
- Observa explainsFacts, gaps, and policy results become readable.
- Operator reviewsA person decides what requires follow-up.
Current release boundary
What this release does, and what it does not
What it does
- Observes supported and mapped OpenClaw activity
- Records supported tool attempts and completions
- Produces metadata-minimized receipts
- Evaluates policy in shadow mode
- Identifies evidence gaps, uncertain results, and review items
- Remains non-blocking when the remote service is unavailable
What it does not do
- Does not actively block or approve actions
- Does not rewrite agent behavior
- Does not guarantee every tool or connected application is represented
- Does not certify compliance, completeness, or safety
- Does not make unmapped activity governable by assumption
“We would rather tell you what we cannot see than pretend we see everything.”
Verification and proof
Restrained claims. Identified release.
The evidence below applies to the verified v0.5.1 public release. It separates McPherson AI release verification, ClawHub availability, and the isolated public-install smoke test.
250 release checks passed0 failures · 0 skipped · McPherson AI verification
Public ClawHub package availableLatest v0.5.1 · clean scan status
Public-install smoke test passedExact ClawHub artifact · isolated named profile
Shadow-only authority boundary verifiedActive enforcement remains false
Managed starting point
Founding Governance Setup
For OpenClaw operators who want help mapping agent activity, defining initial policy, reviewing evidence quality, and identifying coverage gaps before considering enforcement.
Company lanes
Governance leads. Operating experience stays visible.
McPherson Governance
OpenClaw-first policy evaluation, evidence binding, review, approvals, incident recovery, and eventual bounded enforcement.
Built from operations
Restaurant operating workflows developed from sixteen years of direct operating experience.
Narrow, accountable builds
Workflow mapping, source-of-truth discovery, approval design, regulated discovery, and narrow automation builds.